CN101984434B - Webpage data extracting method based on extensible language query - Google Patents
Webpage data extracting method based on extensible language query Download PDFInfo
- Publication number
- CN101984434B CN101984434B CN201010545520A CN201010545520A CN101984434B CN 101984434 B CN101984434 B CN 101984434B CN 201010545520 A CN201010545520 A CN 201010545520A CN 201010545520 A CN201010545520 A CN 201010545520A CN 101984434 B CN101984434 B CN 101984434B
- Authority
- CN
- China
- Prior art keywords
- node
- data
- label
- path
- attribute
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Expired - Fee Related
Links
Images
Landscapes
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
- Information Transfer Between Computers (AREA)
Abstract
A webpage data extracting method based on extensible language query belongs to the technical field of computer database; and the method comprises the following steps: determining the corresponding mode structure in the Web page when extracting the data contents; locating the data area, the data unit and the attribute text in the Web page; marking the semanteme of the attribute text; generating the data unit node path; calculating the path expression form of extracting the attribute value; generating the XML query sentence for extracting the data; and extracting the data by means of the XML query sentence. The method can generate precision XML query sentence for guaranteeing the correctness of the XML query sentence; the method has high generality and can be combined with the current method in seamless; and the method can adapt to more complex query result output.
Description
Technical field
The invention belongs to the computer database technology field, particularly a kind of web data abstracting method based on the extend markup language inquiry.
Background technology
Continuous development along with the Web field; Data message increases rapidly among the Web; Current each application continues to increase the demand of Web data; Though comprised a large amount of structurings and semi-structured data among the Web, these data owners will offer the user through browser with the form of hypertext markup language HTML and browse, and are difficult to directly be used among the application such as data mining and data integration; Therefore how efficiently and exactly from a large amount of Web pages drawing-out structureization become more and more important with semi-structured data, mainly be divided three classes to the typical abstracting method of Web data: based on the method for html tag tree or DOM Document Object Model dom tree; Method based on page structure; Method based on visual information; Method based on html tag tree or DOM Document Object Model dom tree mainly comprises XWRAP, RoadRunner, Lixto, MDR and MDRII etc.; Method groundwork based on page structure comprises NoDoSE, DEByE and SG-WRAP etc.; Method based on visual information is main with ViDRE mainly;
Is a kind of commonplace method based on html tag tree or DOM Document Object Model dom tree to data recording extraction in the page; Before extracted data, be the DOM Document Object Model dom tree with the Web conversion of page at first, then based on architectural feature in the dom tree and automatic or automanual decimation rule extracted data therefrom based on label; Method based on page structure is at first formulated the structure that comprises data division in the page; In the page, seek similar part as extracting the result according to this structure again, still, for the page simple in structure; It can obtain good effect; If in the page dom tree in complex structure and the data field noise node too much, then treatment effect is not fine, but also can't support the data identification of nested structure;
The position habit characteristic of mainly utilizing user's browsing content in the webpage design based on the technology of visual information extracted data in the webpage is extracted data from the relevant position, a kind of abstracting method that the ViDRE of Microsoft Research, Asia proposes based on visualization feature, and this method is simulated the identifying of human eye to the page to a certain extent; Finally reach the purpose of identifying object information; Yet, on the one hand, when the page does not have tangible visual signature; Extraction efficiency based on vision can seriously reduce; And on the other hand, be applicable to based on the mode of vision the single page carried out data pick-up that the page efficiency in extracting different for the identical data of a large amount of structures will be very low;
Above method is only applicable to comprise the webpage of simple data structure, will be difficult to expression or produce the attribute disappearance if the data in the webpage are hierarchical relationship then the result that extracts, therefore is difficult to the complicated content of pages of handle data structures; Secondly, these methods directly generate after initialization and extract result data, if wherein have the Attribute Recognition mistake then to be difficult to timely correction; In addition, these method operations are relatively very independent, are difficult to combine with available data storehouse system, therefore lack the unified management to web data.
Summary of the invention
For remedying the deficiency of said method, the present invention provides a kind of web data abstracting method based on the extensible language inquiry.
Technical scheme of the present invention is achieved in that based on the web data abstracting method of extensible language inquiry, may further comprise the steps:
Step 1: pairing mode configuration when confirming in the Web page extracted data content;
Mode configuration comprises: 2 kinds of the list structure of relation form and hierarchical structures, wherein, the data pattern S of list structure is by data entity name E and one group of community set A={A
1..., A
nConstitute A wherein
iAn attribute in the representation attribute set constitutes 1＜=i＜=n, the quantity of n representation attribute, A by the data type of Property Name and attribute
iBe expressed as<n, Type>, N representation attribute title wherein, Type representation attribute data type, said data type Type comprises integer type integer, floating point type float and character string type string; Described hierarchical structure is meant the complex data structures of being made up of fundamental type, and its corresponding data pattern is expressed as S
i', comprise attribute { A
i'
1..., A
i'
x, x is a Mode S
iThe quantity of ' middle attribute;
Step 2: data area, data cell and attribute text in the Web page of location;
The Web page source code format of html language description is turned to the document of XML language;
Said data area Da is meant the zone that minimum border comprised that in the Web page, comprises all data cells, and localization method is: corresponding minimum subtree that comprises all data cells in the corresponding DOM Document Object Model DOM structure of the Web page;
Said data cell Du, the corresponding data entity of a mode configuration that expression Web data pick-up institute will obtain usually by the attribute description in the pattern, repeats appearance with certain rules in the page; Localization method is: in the DOM Document Object Model dom tree of the Web page, find out the node at each property content place of data entity in the page, the minimum subtree that comprises these nodes is exactly a data cell;
Said attribute text At; Be illustrated in the content of text that comprises the property value of data pattern attribute in the Web page; Usually in the DOM Document Object Model dom tree of the Web page in the text node of node element, localization method is property value: in the corresponding DOM Document Object Model dom tree structure of the Web page, find out the node that comprises this property value text;
Step 3: the attribute text in the step 2 carries out semantic tagger;
Method is: each the attribute text for being comprised in each data cell is all specified the attribute that is comprised in one or more data patterns;
Step 4: generate the data cell node path, may further comprise the steps:
Step 4-1: the data cell set that step 2 is obtained is expressed as: U={U
1, U
2..., U
y, wherein, U
iRepresent a data unit, i=1 wherein ..., y;
Step 4-2: according to established data unit U
i, institute is to deserved node element in page XML document to confirm it, and this node table is shown N
i, the structure according to XML document is node element N again
iThe path values of generation from root node to this node is expressed as P
i
Step 4-3: the path expression of computational data unit, method is:
Get the path of a data cell node, at path values P
iIn; Each step in the predicate location path expression formula of use location; Promptly by the documentation root node to the corresponding node element of data cell each node of process, get each node label in the path expression, the path of all data cells has identical sequence label; The sequence label that then begins from root node is expressed as T, is expressed as (T respectively comprising m label
1, T
2..., T
m), label T wherein
1Be the label of root node, all the other labels and the like, the label of each node is expressed as (p at it with the position sequence in the label brotgher of node
I1..., p
Im), position p wherein
I1Be the position of root node label, all the other labels and the like, then path values is expressed as:
Path values P
i=/label 1 [position i1]/label 2 [position i2]/... / label m [position im],
Be P
i=/T
1[p
I1]/T
2[p
I2]/.../T
m[p
Im]/
Step 4-4:, calculate the longest common path LCP that begins from root node to the set of paths of data unit:
The longest said common path is meant the path that the total node in the path of all data cell nodes constitutes; The method of calculating the longest common path LCP is: for the path of data cell node; First label position that begins from root node begins coupling; If the positional value of all data cell node paths under current label is identical, i.e. p
1i=p
2i=...=p
Yi, then add current label and positional value order in the longest common path to, i.e. LCP+=/T
i[p
i], if there is different value in the positional value of all data cell node paths under current label, then stop the coupling, with the longest current common path value as the longest final common path value;
Step 4-5: the longest common path LCP that abbreviation step 4-4 calculates;
For one in the longest common path pairing node of step, be expressed as n
i, corresponding label is T
iIf, do not exist identically in its brotgher of node with its label, and to have identical successor path be "/label
I+1/ .../label
m" the non-data cell node of descendants's node, then the positional value of this node can omit in the expression formula of the longest common path;
Step 4-6: adopt the method that generates predicate to calculate local path; Described local path is meant the path that the privately owned node of each node constitutes; It is the predicate expression formula of location node accurately, can in all data cell nodes in location, filter incoherent node:
The method that generates predicate is: the label of supposing the node in current step is T
i, see in all brotghers of node of node set in the current step whether comprising identical with its label and having identical successor path is "/label
I+1/ ... ../label
m" the non-data cell node of descendants's node; if then do not omit predicate; if having to check again then whether the XML of non-data cell node attribute is arranged in the present node, can present node and the non-data cell node that meets top condition be distinguished, and if such XML attribute were arranged with this attribute as the predicate expression formula; if there is not then further to calculate the scope of positional value in the predicate, call the noise node to these qualified non-data cell nodes;
The method of the scope of positional value is following in the said calculating predicate:
If the noise node only appears at before the data cell node set, then according to the scope of position in the predicate of cell node be: from the label T of all data cell nodes for this label list registration
iPositional value minimum in the pairing node location is to a last node with this label;
If the noise node only appears at after the data cell node set, then according to the scope of position in the predicate of cell node be: label T from first to all data cell nodes for this label list registration
iThe positional value of maximum in the pairing node location;
If back end is cut apart by the noise node regularly, the interval p that the computational data cell node is cut apart by the noise node
Inte, the length p that the computational data cell node occurs continuously
Cont, and calculate the label T of all data cell nodes
iThe positional value of minimum and maximum positional value are expressed as pmin and pmax in the pairing node location, and the node that then meets the lower position condition is considered to the node on the data cell path: after (1) node location value deducts pmin, to p
InteRemainder behind the delivery is less than p
ContValue; (2) the node location value deducts pmax less than maximum noise node location value and adds the value after 1 again;
Step 4-7: merge the longest common path and local path;
Will the longest common path and local path merge, obtain the path P u of locator data unit in the XML document of the Web page;
Step 5: calculate the path expression that extracts property value, may further comprise the steps;
Step 5-1: the path that generates the attribute node location;
Suppose that in sample data the node at the property value place of pattern attribute Ai with respect to the path representation of data cell node is:
/ label A
I1[position A
I1]/label A
I2[position A
I2]/... / label A
Ik[position A
Ik]
Promptly/TA
I1[pA
I1]/TA
I2[pA
I2]/... / TA
Ik[pA
Ik], TA wherein
IjThe expression label A
Ij, pA
IjExpression position A
Ij, j=1 wherein ..., k, label A
IkFor comprising the label of property value node, position A
IkFor this node at it with the position in the label brotgher of node, then can use the method for step 4-5, abbreviation is carried out in the path of attribute node location;
Step 5-2: confirm the property value decimation rule;
The property value decimation rule is applicable to following two kinds of situation: 1, the property value of a plurality of attributes is contained in the node text simultaneously; 2, comprise non-property value content of text in the node content of text;
Suppose that non-property value content of text is a fixed text in the node text; And also use fixing text to cut apart between the property value of the different attribute in same node text; Only need calculate the property value that the fixed character string of cutting apart attribute in the node text gets final product unbundled attribute value text or different attribute, method is:
At first get a plurality of sample Web pages; Therefrom extract the node text that comprises same alike result; If alphabet is the property value content then directly extraction in this node text, otherwise extracts public substring and cut apart attribute, the rule of extraction property value is as follows from the node text:
If attribute A in the node text
iProperty value before fixed text Text1 arranged, then at first node text-string Str is got the substring Str-after after the fixed text Text1, check attribute A again
iProperty value after, if fixed text Text2 is arranged, then again character string Str-after is got the substring before the fixed text Text2, be expressed as Str-before;
Step 6: the XML query statement that generates data pick-up;
Back end path and attribute node path that the structure of the XML query statement of drawing-out structure data mainly is based on step 4 and step 5 and is obtained; When using the XQuery query language; The structure of statement mainly is to use the FLWOR expression formula of XQuery query language; Wherein, each clause's function is following:
FOR clause: locator data cell node set;
LET clause: increase predicate variable;
WHERE clause: the predicate of data cell node based on the attribute path filtered;
ORDER clause: the rule that the result is sorted;
RETURN clause: return the desired data layout of user;
According to XML query language XQuery syntactic property, can extract the data content of the hierarchical structure in the Web page through nested FLWOR clause's in RETURN clause form, be several kinds of methods that make up the XML query statement according to different demands below:
Step 6-1: when the data pick-up result is hierarchical structure, the XML query statement structure construction method that will generate be:
(1) outermost layer of statement uses fixing XML element tags as root node, and the centre is the XML query expression, is the FLWOR expression formula for the XQuery language, promptly uses following form:<root node Biao Qian>The XML query expression</root node Biao Qian>
(2) in the XML query expression; Use the path expression locator data cell node variable of data cell; Use FOR statement locator data cell node variable for the XQuery language, can use LET statement and WHERE statement to add the predicate of locator data cell node simultaneously;
(3) in the XML query expression; Output at Query Result; Use the attribute-name in the data pattern or have the label of the text of identical semanteme as element in the XML document; Use the path of the attribute node location that generates in the step 5 and the property value decimation rule is located corresponding attribute under the data cell node variable property value text, concrete form is:<shu Xingbiaoqian>{ expression formula that attribute node path and property value decimation rule constitute }</Shu Xingbiaoqian>
The one-piece construction of XML query statement is:
< root node label >
{
FOR data cell node variable in data cell node path
[LET statement]
[WHERE statement]
RETURN < data entity name label >
<attribute 1 Biao Qian>{ expression formula that attribute 1 node path and property value decimation rule constitute }</attribute 1 Biao Qian>
……
<attribute n Biao Qian>{ expression formula that attribute n node path and property value decimation rule constitute }</attribute n Biao Qian>
</>data entity name label;
}
</root node Biao Qian>
Step 6-2: when the data pick-up result is the list structure of relation form, the XML query statement structure construction method that will generate:
1 in the XML query expression; Use the path expression locator data cell node variable of data cell; Use FOR statement locator data cell node variable for the XQuery language, use LET statement and WHERE statement to add the predicate of locator data cell node simultaneously;
2 in the XML query expression; Output at Query Result; Demand according to the output result; Be arranged in order the expression formula that is made up of attribute node path and property value decimation rule, separate with special symbol between the expression formula of different attribute value, concrete form is: { property value of attribute 1 extracts expression formula } separator { property value of attribute 2 extracts expression formula } separator ... Separator { property value of attribute n extracts expression formula }
Step 7: utilize XML query statement extracted data;
Use the execution engine of XML query processing, operation XML query statement can extract the data designated content from the webpage that is formatted as the XML document form on the XML document after the target web conversion.
Advantage of the present invention: the Web data pick-up method based on the XML inquiry of the present invention has than extensive applicability: (1) the present invention can generate accurate XML query statement; Based on path expression generation method; Data unit and property value are carried out accurate XPath expression formula location, thereby guarantee the correctness of XML query statement; (2) the present invention has high generality, and the XML query statement of data source extracted data may operate in database in generation or the XML query specification is carried out on the engine, can with existing seamless fusion; (3) the present invention can adapt to complicated query result output more, through the structure of adjustment bearing-age tree virgin sentence, supports to extract the data content of the middle-level structure of the Web page, not only is confined to simple relational structure.
Description of drawings
Fig. 1 is that the web data abstracting method electronics that the present invention is based on the extensible language inquiry is sold the Web page of data synoptic diagram of book website;
Fig. 2 is the web data abstracting method process flow diagram that the present invention is based on the extensible language inquiry;
Fig. 3 the present invention is based on the position view of the web data abstracting method data cell of extensible language inquiry at the page documents dom tree.
Embodiment
Below in conjunction with accompanying drawing and embodiment the present invention is done further explain:
Fig. 1 sells a Web page of data of book website for certain electronics, adopts the flow process of the inventive method as shown in Figure 2, and step is following:
Step 1: pairing data pattern S when confirming from the Web page extracted data content, wherein data entity title E is " books ", the Property Name that community set comprised and the data type of attribute are as shown in table 1:
Table 1 is data entity " books " Property Name that is comprised and the data type of attribute
Step 2: the data area in the location map 1 in the sample page, data cell and attribute text, can know that from Fig. 1 data cell is made up of data cell 1, data cell 2, data cell 3;
The XML document that at first need being formatted as of html page be met the XML language standard:
<div?class＝″list_book_right″>
<h2><img/><a name=" link_prd_name " href=" " target=" _ blank ">Algorithm and the data structure examination question essence of preparing for the postgraduate qualifying examination is analysed (the 2nd edition)</a></h2>
<h3>Client's scoring:</h3>
<h4 class=" list_r_list_h4 ">The author:<a href=" ">Chen Shoukong</a>,<a href=" ">Hu Xiaokun</a>,<a href=" ">Li Ling</a>Write</h4>
<h4>Publishing house:<a href=" ">China Machine Press</a></h4>
<h4>Publication time: 2007 07 month</h4>
< h5>this book collected key university of institute and academy of sciences surplus in the of since nineteen ninety-two domestic 60, the 1600 multiple tracks examination questions of more than 300 cover Master degree candidates entrance " algorithm and data structure " examination papers, and provided Key for Reference and analysis.This book can be used as institution of higher learning's computing machine and relevant speciality learning data<font class=dot>...</font></h5>
<div?class＝″clear″></div>
<h6><span>$42.00</span><span>$35.70</span>Discount: 85 foldings are saved: $6.30</h6>
<span?class＝″list_r_list_button″><a?href＝”><img?src＝”/></a></span>
<span?class＝″list_r_list_button″><a?href＝″″><img?src＝″″/></a></span>
</div>
Step 3: the attribute text to data cell in the sample page marks, and the text among Fig. 1 in 3 data unit marks respectively as follows:
Data cell 1:
Title: algorithm and the data structure examination question essence of preparing for the postgraduate qualifying examination is analysed (the 2nd edition)
Author: Chen Shoukong, Hu Xiaokun, Li Ling
Publishing house: China Machine Press
Publication time: 2007 07 month
The books brief introduction: this book collected key university of institute and academy of sciences surplus in the of since nineteen ninety-two domestic 60, the 1600 multiple tracks examination questions of more than 300 cover Master degree candidates entrance " algorithm and data structure " examination papers; And having provided Key for Reference and analysis, this book can be used as institution of higher learning's computing machine and relevant speciality learning data
Original cost: $42.00
Present price: $35.70
Discount: 85
Save amount of money: $6.30
Data cell 2:
Title: data mining notion and technology (former book the 2nd edition)
Author: Han Jiawei may win, and model is bright, Meng Xiaofeng
Publishing house: China Machine Press
Publication time: 2007 03 month
The books brief introduction: this book is told about the important knowledge and technology innovation in data mining field all sidedly; On the quite comprehensive basis of the 1st version content; The 2nd edition newest research results of having showed this field; For example excavate stream, sequential and sequence data and excavate time and space, multimedia, text and web data, this book can be used as
Original cost: $55.00
Present price: $42.30
Discount: 77
Save amount of money: $12.70
Data cell 3:
Title: the Oracle9i&10g art of programming: go deep into data base architecture
Author: Kate, Su Jinguo
Publishing house: People's Telecon Publishing House
Publication time: in October, 2006
The books brief introduction: this book is the authoritative books about oracle 9j az&10g data base architecture, has contained all most important oracle architecture characteristicses, comprises file, internal storage structure and process; Lock and door bolt; Affairs, concurrent and many versions, table and index, data type; And subregion and parallel, and
Original cost: $99.00
Present price: $74.30
Discount: 75
Save amount of money: $24.70
Step 4: the path that generates the data cell node;
Step 4-1: data cell U={U1; U2; U3}, wherein: U1 representes that data cell 1, U2 represent that data cell 2, U3 represent data cell 3, wherein the title of data cell 1 is " algorithm and data structure prepare for the postgraduate qualifying examination examination question essence analyse (the 2nd edition) "; The title of data cell 2 is " a data mining notion and technology (former book the 2nd edition) ", and the title of data cell 3 is " the Oracle9i&10g art of programming: go deep into data base architecture ";
Step 4-2: data cell marks out in Fig. 1; Position in the data cell corresponding page document D OM tree shown in solid dot among Fig. 3, among Fig. 3 in the corresponding XML document of root node label be the element of html, the outermost layer dotted line of Webpage in the corresponding diagram 1; Comprising whole content viewables and not visual content; Node 1 is the XML node element of head for label, the web data header in the corresponding diagram 1 in the Webpage, and wherein the page metamessage that comprises of content is not visual element; Node 2 is the XML node element of body for label; Outermost layer solid line in the corresponding diagram 1 in the Webpage,, node 2.1 to node 2.7 all is that label is the XML node element of div in the child nodes of node 2; The below advertisement position in the Webpage in node 2.7 corresponding diagram 1 wherein; Triangle among the figure under the node is represented the subtree under this node, and node 2.6.1 is that label is the XML node element of div in the child nodes of node 2.6 to node 2.6.3, wherein solid line zone pointed, the data area of Webpage in the node 2.6.3 corresponding diagram 1; Node 2.6.3.1 is that first label is the XML node element of div in the node 2.6.3 child nodes; The data cell 1 of Webpage solid line zone pointed in the node 2.6.3.1 corresponding diagram 1 wherein, node div [1] representes that this node is that first label is the XML node element of div in its father node child nodes, the pairing node in solid node bit data unit among Fig. 3; The path of this node is exactly the data cell path, comprises the attribute value data of data cell in the text node of this node subtree;
Step 4-3: calculating path expression formula;
The path values of the node of data cell in XML document is respectively:
P1：“/html[1]/body[1]/div[6]/div[3]/div[1]/div[4]/div[2]”
P2：“/html[1]/body[1]/div[6]/div[3]/div[1]/div[5]/div[2]”
P3：“/html[1]/body[1]/div[6]/div[3]/div[1]/div[6]/div[2]”
Html wherein, body, div are the XML element tags, div [i] expression label is that the div node is i in its brotgher of node with label;
Step 4-4: calculate the longest common path LCP that begins from root node;
The longest common path LCP:
LCP：“/html[1]/body[1]/div[6]/div[3]/div[1]”
Step 4-5: the longest common path LCP that abbreviation step 4-3 calculates;
The longest common path expression formula behind the abbreviation is:
LCP：“/html/body/div[6]/div[3]/div[1]”
Step 4-6: calculate local path;
The expression formula of local path is:
“/div[.class＝″list_r_list″]/div[2]”
Step 4-7: merge the longest common path and local path;
The path expression that obtains the data cell node after the merging is:
“/html/body/div[6]/div[3]/div[1]/div[.class＝″list_r_list″]/div[2]”
Wherein div [.class=" list_r_list "] expression has XML attribute class, and Class is the Style Attributes of label node in the html document, and property value is that the XML label of list_r_list is the node element of div;
Step 5: generate the path expression that extracts property value;
For data cell 1, it is following to generate the path expression that extracts property value based on structure wherein:
1. attribute " title "
Path localization and expression formula is "/h2/a/ ", and wherein h2 representes that label is the XML node element of h2, and a representes that label is the XML node element of a;
The property value decimation rule: the content in this attribute node all is a property value information, can use text () function in the XQuery language as the function that extracts property value.
2. attribute " author "
Path localization and expression formula is "/h4 [1]/a ", and wherein h4 [1] representes that first label is the XML node element of h4, and a representes that label is the XML node element of a;
The property value decimation rule: the content in this attribute node all is a property value information.
3. attribute " publishing house "
Path localization and expression formula is "/h4 [2]/a ", and wherein h4 [2] representes second XML node element that label is h4, and a representes that label is the XML node element of a;
The property value decimation rule: the content in this attribute node all is a property value information.
4. attribute " publication time "
Path localization and expression formula is "/h4 [3] ", wherein the 3rd XML node element that label is h4 of h4 [3] expression;
The property value decimation rule: the content part in this attribute node is a property value information, and decimation rule is for eliminating public non-property value character string " publication time: ".
5. attribute " books brief introduction "
Path localization and expression formula is "/h5 ", and wherein h5 representes that label is the XML node element of h5;
The property value decimation rule: the content in this attribute node all is a property value information.
6. attribute " original cost "
Path localization and expression formula is "/h6/span [1] ", and wherein h6 representes that label is the XML node element of h6, and span [1] representes that first label is the XML node element of span;
The property value decimation rule: the content part in this attribute node is a property value information, and decimation rule is for eliminating public non-property value character string " $ ".
7. attribute " present price "
Path localization and expression formula is "/h6/span [2] ", and wherein h6 representes that label is the XML node element of h6, and span [2] representes second XML node element that label is span;
The property value decimation rule: the content part in this attribute node is a property value information, and decimation rule is for eliminating public non-property value character string " $ ".
8. attribute " discount "
Path localization and expression formula is "/h6/ ", and wherein h6 representes that label is the XML node element of h6;
The property value decimation rule: the content part in this attribute node is a property value information, and decimation rule is for eliminating node text-string public non-property value character string of middle front part " discount: " and the public non-property value character string in rear portion " folding ".
9. attribute " the saving amount of money "
Path localization and expression formula is "/h6/ ", and wherein h6 representes that label is the XML node element of h6;
The property value decimation rule: the content part in this attribute node is a property value information, and decimation rule " is saved: $ " for eliminating the public non-property value character string of node text-string middle front part.
Step 6: the XML query statement that generates data pick-up;
To generate XML form extraction result data is example, uses XQuery as the XML query language, following for the XQuery statement that this example generated:
Wherein, FOR, IN, RETURN are XQuery query language key word; Text () is for obtaining the function of intranodal text; Substring-before () is for obtaining the substring function before a certain special string in the character string, and substring-after () is for obtaining the substring function after a certain special string in the character string.
Step 7: carry out above XML query statement extracted data;
Behind the data pick-up of above XML query statement execution to example page, the XML data content of acquisition is:
<books tabulation >
<books >
<shu Ming>Algorithm and the data structure examination question essence of preparing for the postgraduate qualifying examination is analysed (the 2nd edition)</Shu Ming>
<zuo Zhe>Oldly keep Kong Huxiao a kind of jade Li Ling</Zuo Zhe>
<chu Banshe>China Machine Press</Chu Banshe>
<chu Banshijian>2007 07 month</Chu Banshijian>
<books brief introduction>this book collected key university of institute and academy of sciences surplus in the of since nineteen ninety-two domestic 60, the 1600 multiple tracks examination questions of more than 300 cover Master degree candidates entrance " algorithm and data structure " examination papers, and provided Key for Reference and analysis.This book can be used as institution of higher learning's computing machine and relevant speciality learning data</books Jian Jie>
<yuan Jia>42.00</Yuan Jia>
<xian Jia>35.70</Xian Jia>
<zhe Kou>85</Zhe Kou>
<save Jin E>6.30</saving Jin E>
</Tu Shu>
<books >
<shu Ming>Data mining notion and technology (former book the 2nd edition)</Shu Ming>
<zuo Zhe>Han Jiawei may the Bo Fanmingmeng small peak</Zuo Zhe>
<chu Banshe>China Machine Press</Chu Banshe>
<chu Banshijian>2007 03 month</Chu Banshijian>
<books brief introduction>this book is told about the important knowledge and technology innovation in data mining field all sidedly.On the quite comprehensive basis of the 1st version content, the 2nd edition newest research results of having showed this field for example excavated stream, sequential and sequence data and excavated time and space, multimedia, text and web data.This book can be used as</books Jian Jie>
<yuan Jia>55.00</Yuan Jia>
<xian Jia>42.30</Xian Jia>
<zhe Kou>77</Zhe Kou>
<save Jin E>12.70</saving Jin E>
</Tu Shu>
<books >
<shu Ming>The Oracle9i&10g art of programming: go deep into data base architecture</Shu Ming>
<zuo Zhe>The Kate Jin nation of reviving</Zuo Zhe>
<chu Banshe>The People's Telecon Publishing House</Chu Banshe>
<chu Banshijian>In October, 2006</Chu Banshijian>
<books Jian Jie>This book is the authoritative books about oracle 9j az&10g data base architecture, has contained all most important oracle architecture characteristicses, comprises file, internal storage structure and process; Lock and door bolt, affairs, concurrent and many versions, table and index; Data type, and subregion and parallel, and</books Jian Jie>
<yuan Jia>99.00</Yuan Jia>
<xian Jia>74.30</Xian Jia>
<zhe Kou>75</Zhe Kou>
<save Jin E>24.70</saving Jin E>
</Tu Shu>
<books tabulation >
Claims (1)
1. web data abstracting method based on extend markup language inquiry is characterized in that: may further comprise the steps:
Step 1: pairing mode configuration when confirming in the Web page extracted data content;
Step 2: data area, data cell and attribute text in the Web page of location;
Step 3: the attribute text in the step 2 carries out semantic tagger;
Step 4: generate the data cell node path;
Step 5: calculate the path expression that extracts property value;
Step 6: the XML query statement that generates data pick-up;
Step 7: utilize XML query statement extracted data;
Wherein, the described mode configuration of step 1 comprises: two kinds of the list structure of relation form and hierarchical structures, wherein, the data pattern S of list structure is by data entity name E and one group of community set A={A
1..., A
nConstitute A wherein
iAn attribute in the representation attribute set constitutes 1＜=i＜=n, the quantity of n representation attribute, A by the data type of Property Name and attribute
iBe expressed as<n, Type>, N representation attribute title wherein, Type representation attribute data type, said data type Type comprises integer type integer, floating point type float and character string type string; Described hierarchical structure is meant the complex data structures of being made up of fundamental type, and its corresponding data pattern is expressed as S
i', comprise attribute { A
i'
1..., A
i'
x, x is a Mode S
iThe quantity of ' middle attribute;
The said data area Da of step 2 is meant the zone that minimum border comprised that in the Web page, comprises all data cells, and localization method is: corresponding minimum subtree that comprises all data cells in the corresponding DOM Document Object Model DOM structure of the Web page;
Said data cell Du, the corresponding data entity of a mode configuration that expression Web data pick-up institute will obtain by the attribute description in the pattern, repeats appearance with certain rules in the page; Localization method is: in the DOM Document Object Model dom tree of the Web page, find out the node at each property content place of data entity in the page, the minimum subtree that comprises these nodes is exactly a data cell;
Said attribute text At; Be illustrated in the content of text that comprises the property value of data pattern attribute in the Web page; In the text node of property value node element in the DOM Document Object Model dom tree of the Web page, localization method is: in the corresponding DOM Document Object Model dom tree structure of the Web page, find out the node that comprises this property value text;
The said generation data cell of step 4 node path may further comprise the steps:
Step 4-1: the data cell set that step 2 is obtained is expressed as: U={U
1, U
2..., U
y, wherein, U
iRepresent a data unit, i=1 wherein ..., y;
Step 4-2: according to established data unit U
i, institute is to deserved node element in page XML document to confirm it, and this node table is shown N
i, the structure according to XML document is node element N again
iThe path values of generation from root node to this node is expressed as P
i
Step 4-3: the path expression of computational data unit, method is:
Get the path of a data cell node, at path values P
iIn; Each step in the predicate location path expression formula of use location; Promptly by the documentation root node to the corresponding node element of data cell each node of process, get each node label in the path expression, the path of all data cells has identical sequence label; The sequence label that then begins from root node is expressed as T, is expressed as (T respectively comprising m label
1, T
2..., T
m), label T wherein
1Be the label of root node, all the other labels and the like, the label of each node is expressed as (p at it with the position sequence in the label brotgher of node
I1, p
Im), position p wherein
I1Be the position of root node label, all the other labels and the like, then path values is expressed as:
Path values P
i=/label 1 [position i1]/label 2 [position i2]/... / label m [position im],
Be P
i=/T
1[p
I1]/T
2[p
I2]/.../T
m[p
Im]/
Step 4-4:, calculate the longest common path LCP that begins from root node to the set of paths of data unit:
The longest said common path is meant the path that the total node in the path of all data cell nodes constitutes; The method of calculating the longest common path LCP is: for the path of data cell node; First label position that begins from root node begins coupling; If the positional value of all data cell node paths under current label is identical, i.e. p
1i=p
2i=...=p
Yi, then add current label and positional value order in the longest common path to, i.e. LCP+=/T
i[p
i], if there is different value in the positional value of all data cell node paths under current label, then stop the coupling, with the longest current common path value as the longest final common path value;
Step 4-5: the longest common path LCP that abbreviation step 4-4 calculates;
For one in the longest common path pairing node of step, be expressed as n
i, corresponding label is T
iIf, do not exist identically in its brotgher of node with its label, and to have identical successor path be "/label
I+1/ .../label
m" the non-data cell node of descendants's node, then the positional value of this node can omit in the expression formula of the longest common path;
Step 4-6: adopt the method that generates predicate to calculate local path, described local path is meant the path that the privately owned node of each node constitutes:
The method that generates predicate is: the label of supposing the node in current step is T
i, see in all brotghers of node of node set in the current step whether comprising identical with its label and having identical successor path is "/label
I+1/ .../label
m" the non-data cell node of descendants's node; if then do not omit predicate; if having to check again then whether the XML of non-data cell node attribute is arranged in the present node, can present node and the non-data cell node that meets top condition be distinguished, and if such XML attribute were arranged with this attribute as the predicate expression formula; if there is not then further to calculate the scope of positional value in the predicate, call the noise node to these qualified non-data cell nodes;
The method of the scope of positional value is following in the said calculating predicate:
If the noise node only appears at before the data cell node set, then for this label list registration according to the scope of position in the predicate of cell node be: positional value minimum from the pairing node location of label Ti of all data cell nodes is to a last node with this label;
If the noise node only appears at after the data cell node set, then according to the scope of position in the predicate of cell node be: label T from first to all data cell nodes for this label list registration
iThe positional value of maximum in the pairing node location;
If back end is cut apart by the noise node regularly, the interval p that the computational data cell node is cut apart by the noise node
Inte, the length p that the computational data cell node occurs continuously
Cont, and calculate the label T of all data cell nodes
iThe positional value of minimum and maximum positional value are expressed as pmin and pmax in the pairing node location, and the node that then meets the lower position condition is considered to the node on the data cell path: after (1) node location value deducts pmin, to p
InteRemainder behind the delivery is less than p
ContValue; (2) the node location value deducts pmax less than maximum noise node location value and adds the value after 1 again;
Step 4-7: merge the longest common path and local path;
Will the longest common path and local path merge, obtain the path P u of locator data unit in the XML document of the Web page;
The path expression that property value is extracted in the described calculating of step 5 may further comprise the steps:
Step 5-1: the path that generates the attribute node location;
Suppose in sample data pattern attribute A
iThe node at property value place with respect to the path representation of data cell node be:
/ label A
I1[position A
I1]/label A
I2[position A
I2]/... / label A
Ik[position A
Ik]
Promptly/TA
I1[pA
I1]/TA
I2[pA
I2]/... / TA
Ik[pA
Ik], TA wherein
IjThe expression label A
Ij, pA
IjExpression position A
Ij, j=1 wherein ..., k, label A
IkFor comprising the label of property value node, position A
IkFor this node at it with the position in the label brotgher of node, then can use the method for step 4-5, abbreviation is carried out in the path of attribute node location;
Step 5-2: confirm the property value decimation rule;
1), the property value of a plurality of attributes is contained in the node text simultaneously the property value decimation rule is applicable to following two kinds of situation:; 2), comprise non-property value content of text in the node content of text;
Suppose that non-property value content of text is a fixed text in the node text; And also use fixing text to cut apart between the property value of the different attribute in same node text; Only need calculate the property value that the fixed character string of cutting apart attribute in the node text gets final product unbundled attribute value text or different attribute, method is:
At first get a plurality of sample Web pages; Therefrom extract the node text that comprises same alike result; If alphabet is the property value content then directly extraction in this node text, otherwise extracts public substring and cut apart attribute, the rule of extraction property value is as follows from the node text:
If attribute A in the node text
iProperty value before fixed text Text1 arranged, then at first node text-string Str is got the substring Str-after after the fixed text Text1, check attribute A again
iProperty value after, if fixed text Text2 is arranged, then again character string Str-after is got the substring before the fixed text Text2, be expressed as Str-before;
The XML query statement of the described generation data pick-up of step 6 may further comprise the steps:
Step 6-1: when the data pick-up result is hierarchical structure, the XML query statement structure construction method that will generate be:
(1) outermost layer of statement uses fixing XML element tags as root node, and the centre is the XML query expression, is the FLWOR expression formula for the XQuery language, promptly uses following form:<root node Biao Qian>The XML query expression</root node Biao Qian>
(2) in the XML query expression; Use the path expression locator data cell node variable of data cell; Use FOR statement locator data cell node variable for the XQuery language, can use LET statement and WHERE statement to add the predicate of locator data cell node simultaneously;
(3) in the XML query expression; Output at Query Result; Use the attribute-name in the data pattern or have the label of the text of identical semanteme as element in the XML document; Use the path of the attribute node location that generates in the step 5 and the property value decimation rule is located corresponding attribute under the data cell node variable property value text, concrete form is:<shu Xingbiaoqian>{ expression formula that attribute node path and property value decimation rule constitute }</Shu Xingbiaoqian>
Step 6-2: when the data pick-up result is the list structure of relation form, the XML query statement structure construction method that will generate:
(1) in the XML query expression; Use the path expression locator data cell node variable of data cell; Use FOR statement locator data cell node variable for the XQuery language, use LET statement and WHERE statement to add the predicate of locator data cell node simultaneously;
(2) in the XML query expression; Output at Query Result; Demand according to the output result; Be arranged in order the expression formula that is made up of attribute node path and property value decimation rule, separate with special symbol between the expression formula of different attribute value, concrete form is: { property value of attribute 1 extracts expression formula } separator { property value of attribute 2 extracts expression formula } separator ... Separator { property value of attribute n extracts expression formula }.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201010545520A CN101984434B (en) | 2010-11-16 | 2010-11-16 | Webpage data extracting method based on extensible language query |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201010545520A CN101984434B (en) | 2010-11-16 | 2010-11-16 | Webpage data extracting method based on extensible language query |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN101984434A CN101984434A (en) | 2011-03-09 |
| CN101984434B true CN101984434B (en) | 2012-09-05 |
Family
ID=43641603
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201010545520A Expired - Fee Related CN101984434B (en) | 2010-11-16 | 2010-11-16 | Webpage data extracting method based on extensible language query |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN101984434B (en) |
Families Citing this family (19)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102456053B (en) * | 2010-11-02 | 2013-08-14 | 江苏大学 | Method for mapping XML document to database |
| CN102902723A (en) * | 2012-09-06 | 2013-01-30 | 北京北森测评技术有限公司 | Method and device for analyzing network data |
| CN103778104B (en) * | 2012-10-22 | 2017-05-03 | 富士通株式会社 | Information processing device, information processing method and electronic device |
| CN103186674A (en) * | 2013-04-02 | 2013-07-03 | 浪潮电子信息产业股份有限公司 | Web data quick inquiry method based on extensive makeup language (XML) |
| CN107430600A (en) * | 2014-12-12 | 2017-12-01 | 慧与发展有限责任合伙企业 | Expansible web data extraction |
| CN105808520B (en) * | 2014-12-30 | 2018-12-14 | 联想(北京)有限公司 | Electronic equipment and its sentence processing method |
| CN106980619B (en) * | 2016-01-18 | 2021-03-26 | 北京国双科技有限公司 | Data query method and device |
| CN106294722B (en) * | 2016-08-09 | 2019-11-22 | 上海资誉网络科技有限公司 | A kind of web page contents extraction method and device |
| CN107957909B (en) * | 2016-10-17 | 2022-01-07 | 腾讯科技（深圳）有限公司 | Information processing method, terminal equipment and server |
| CN106649628B (en) * | 2016-12-06 | 2020-08-25 | 北京大学 | Interaction enhancement method and system for webpage visualization area |
| CN108614842B (en) * | 2016-12-13 | 2021-03-30 | 北京国双科技有限公司 | Method and device for querying data |
| CN106951451B (en) * | 2017-02-22 | 2019-11-12 | 麒麟合盛网络技术股份有限公司 | A kind of webpage content extracting method, device and calculate equipment |
| CN108334560B (en) * | 2018-01-03 | 2022-04-15 | 腾讯科技（深圳）有限公司 | Information acquisition method and related equipment |
| CN110309364B (en) * | 2018-03-02 | 2023-03-28 | 腾讯科技（深圳）有限公司 | Information extraction method and device |
| CN109582886B (en) * | 2018-11-02 | 2022-05-10 | 北京字节跳动网络技术有限公司 | Page content extraction method, template generation method and device, medium and equipment |
| CN112528082B (en) * | 2020-12-08 | 2022-05-03 | 集美大学 | XML document production line XPath query method, terminal equipment and storage medium |
| CN112836063B (en) * | 2021-01-27 | 2023-06-06 | 四川新网银行股份有限公司 | Method for realizing feature tracing |
| CN114817639B (en) * | 2022-05-18 | 2024-05-10 | 山东大学 | Webpage diagram convolution document ordering method and system based on contrast learning |
| CN115658993B (en) * | 2022-09-27 | 2023-06-06 | 观澜网络(杭州)有限公司 | Intelligent extraction method and system for core content of webpage |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101290624A (en) * | 2008-06-11 | 2008-10-22 | 华东师范大学 | News web page metadata automatic extraction method |
| CN101582074A (en) * | 2009-01-21 | 2009-11-18 | 东北大学 | Method for extracting data of DeepWeb response webpage |
-
2010
- 2010-11-16 CN CN201010545520A patent/CN101984434B/en not_active Expired - Fee Related
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101290624A (en) * | 2008-06-11 | 2008-10-22 | 华东师范大学 | News web page metadata automatic extraction method |
| CN101582074A (en) * | 2009-01-21 | 2009-11-18 | 东北大学 | Method for extracting data of DeepWeb response webpage |
Non-Patent Citations (5)
| Title |
|---|
| 周津等.基于XML的网页信息自动抽取.《计算机应用》.2004,第24卷第225-227页. * |
| 孙高尚等.一种应用于Deep Web结果页面中分页标签的识别方法.《小型微型计算机系统》.2010,第31卷(第4期),第635-640页. * |
| 李剑波等.一种基于XML的Web信息抽取方法.《情报杂志》.2006,(第8期),第49-51页. * |
| 申德荣等.支持Web深层数据库网络的部分关键技术的研究.《计算机科学》.2007,第34卷(第8期),第123-125页. * |
| 邓丽.面向主题的XML网页的模式和数据抽取.《中国优秀硕士学位论文全文数据库信息科技辑》.2004,第1-47页. * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN101984434A (en) | 2011-03-09 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN101984434B (en) | Webpage data extracting method based on extensible language query | |
| Liu et al. | Vide: A vision-based approach for deep web data extraction | |
| Wang et al. | Data-rich section extraction from html pages | |
| Day et al. | Reference metadata extraction using a hierarchical knowledge representation framework | |
| Zheng et al. | Template-independent news extraction based on visual consistency | |
| JP4956757B2 (en) | Formula description structured language object search system and search method | |
| CN103294781B (en) | A kind of method and apparatus for processing page data | |
| US20100030752A1 (en) | System, methods and applications for structured document indexing | |
| CN102662969B (en) | Internet information object positioning method based on webpage structure semantic meaning | |
| Xue et al. | Web page title extraction and its application | |
| CN104142985A (en) | Semi-automatic vertical crawler generation tool and method | |
| Zheng et al. | Efficient record-level wrapper induction | |
| Cruz et al. | Semantic extraction of geographic data from web tables for big data integration | |
| Liu et al. | Automatically extracting user reviews from forum sites | |
| He et al. | Application of the indent conversion based on XML and DOM | |
| Ciardelli et al. | The EDIT cyberplatform for taxonomy and the taxonomic workflow: Selected components | |
| Zhang et al. | Exploiting multi-category characteristics and unified framework to extract web content | |
| Lam et al. | A method for web information extraction | |
| Kolkur et al. | Web Data Extraction Using Tree Structure Algorithms-A Comparison | |
| Kit et al. | OLAP query processing for XML data in RDBMS | |
| Du et al. | Automate discovery of deep web interfaces | |
| Cafarella | Extracting and managing structured web data | |
| Barzilay | Graph-based Algorithms in NLP | |
| Guan et al. | Nstar: an interactive tool for local web search | |
| Chen et al. | An efficient content extraction method for webpage based on tag-line-block analysis |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| C14 | Grant of patent or utility model | ||
| GR01 | Patent grant | ||
| CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20120905 Termination date: 20141116 |
|
| EXPY | Termination of patent right or utility model |